UC Berkeley

The goal of conservation biology is to understand and prevent the loss of biological diversity. Modern conservation science relies heavily on four major quantitative methods: reserve site selection algorithms, species distribution models, population viability analyses, and species-area relationships. These methods, however, have several longstanding and unresolved shortcomings, including extensive data requirements, long computation times, and important simplifying assumptions, that limit their ability to inform conservation decisions in many real landscapes. This dissertation develops new approaches in quantitative ecology that address these shortcomings through the use of simulation modeling, probability theory, machine learning, modern statistics, and economic input-output analysis.

Chapter 2 examines the optimal design of reserve networks for preventing extinction of terrestrial mammal species, demonstrating that the match between a species' body size and the spatial scale of a landscape can be used to determine which species will benefit from a clustered reserve network design. Chapter 3 derives two new macroecological metrics, similar to the species-area relationship, that provide probabilistic estimates of single-species extinction risks and community-wide extinction rates in landscapes undergoing habitat loss. Chapter 4 uses data from acoustic surveys to examine the road ecology of California bats, finding that total bat activity, and the activity of four common species, is decreased in the vicinity of large highways. Chapter 5 presents a "wildlife footprint" analysis that combines global bird and mammal range maps, data on the human appropriation of net primary productivity, and economic input-output tables to link specific human consumption activities to decreases in wild bird and mammal populations.